Will the internet polling argument ever be resolved?

In a move not designed to win him any friends within the UK polling industry the boss of YouGov, Peter Kellner, has posted a strongly argued piece, under the provocative heading “YouGov – consistently right” to reinforce his claim that his firm’s approach to taking the nation’s political temperature is the right one.

Although he acknowledges that all the polls were within the same area in their eve of poll prediction and NOP got it spot on “…the difference is that YouGov, alone, consistently published figures throughout the campaign that were close to the result.

Polls, Kellner says, should be judged not only against their final election-day prediction, but against their results throughout the campaign.

And here he points to the fifteen polls published by the company during the campaign fourteen of which got the Labour share to within one point. This was in sharp contrast, he says, to the others of which the telephone pollsters …during the five weeks leading up to polling day were far less consistent. They ranged from a Conservative lead of five points to a Labour lead of 14 points.

There’s little doubt that the consistent overstatement of Labour’s position by ICM, Populus, Communicate Research and NOP had a big impact on the betting markets. In their final surveys before the ones for May 5th itself ICM reported an 8% Labour lead; NOP a 10% one; CR had 8% and Populus 9% – figures which led to projections of 130-140 seat majorities for Labour and which over-stated the winning vote share margin by 2-3 times what actually happened.

Although the spread markets did not get into the 100+ majority area they did get it above 90 in the closing period. Right from their inception the vote share markets over-stated Labour and everybody who sold came out with a profit.

I am indebted to Anthony Wells of UK Polling Report for maintaining and excellent record of each of the surveys and for making it easier to look at each firm’s record.

Mike Smithson